모델 목록으로
DeepSeek오픈소스

DeepSeek V4 Pro

DeepSeek V4 Pro is a large-scale reasoning model developed by DeepSeek-AI. It has a gigantic scale with approximately 16 trillion parameters and an extensive context window of 1 million tokens.

파라미터

16000.0B

컨텍스트

1M

라이선스

MIT

출시일

2026-04-24

API 가격

입력 가격 (1M 토큰당)

$0.435

출력 가격 (1M 토큰당)

$0.87

과금 모드: standard

강점

  • Overwhelming 16-trillion parameter count
  • Long-context understanding of 1M tokens
  • High openness via MIT license

약점

  • Computational load from huge model size
  • Increased resource consumption during inference
  • Expected high operational costs

활용 사례

  • Extremely complex logical reasoning tasks
  • Analysis of vast documents
  • Research requiring advanced specialized knowledge

심층 분석

Artificial Analysis Intelligence Index

52

#2 open-weights reasoning model, behind Kimi K2.6 (54)

SWE-Bench Verified

80.6%

Ties Claude Opus 4.6 (80.8%) and Gemini 3.1 Pro (80.6%)

LiveCodeBench Pass@1

93.5

#1 among all tested models, ahead of Gemini 3.1 Pro (91.7)

GPQA Diamond

90.1%

Competitive with GPT-5.4 (93.0%) and Gemini 3.1 Pro (94.3%)

Output Price

$3.48/1M (list)

~7x cheaper than Claude Opus 4.7 ($25/1M)

Context Window

1M tokens

8x expansion over V3.2 (128K)

Total Parameters

1.6T (49B active)

MoE architecture; largest open-weights model to date

강점

  • Leads all public models on competitive coding benchmarks (LiveCodeBench 93.5, Codeforces 3206)
  • MIT-licensed open weights enable self-hosting, fine-tuning, and full compliance control
  • 7-8x cheaper output pricing than Claude Opus 4.7 at near-identical SWE-Bench scores

약점

  • Extremely high hallucination rate of 94% — nearly always responds even when uncertain
  • Political censorship embedded in training weights, not just API-level filtering
  • Meaningful knowledge gap on SimpleQA (57.9%) vs Gemini 3.1 Pro (75.6%) and HLE (37.7% vs 44.4%)

경쟁사 비교

ModelSWEGPQAPrice
Claude Opus 4.780.9%94.2%$5.00/$25.00
GPT-5.480.0%93.0%$2.50/$15.00
Gemini 3.1 Pro80.6%94.3%$2.00/$12.00

DeepSeek V4 Pro, released April 24, 2026, is DeepSeek-AI's flagship open-weights reasoning model and the largest MIT-licensed language model to date at 1.6 trillion total parameters (49 billion active via Mixture-of-Experts routing). It introduces a new V4 architecture with a 1-million-token context window, a hybrid attention mechanism combining Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA), and dual thinking/non-thinking inference modes. The model represents a significant generational leap from V3.2, scoring 52 on the Artificial Analysis Intelligence Index (a 10-point gain) and tying or leading closed frontier models on key coding benchmarks.

V4 Pro's core value proposition is delivering near-frontier performance at a fraction of closed-model pricing. At list prices of $1.74/$3.48 per million input/output tokens, it costs roughly one-seventh of Claude Opus 4.7's output pricing while matching it on SWE-Bench Verified (80.6% vs 80.8%) and leading on competitive programming (LiveCodeBench 93.5, Codeforces rating 3206). The efficiency gains come from architectural innovations: at 1M-token context, V4 Pro requires only 27% of V3.2's single-token inference FLOPs and 10% of its KV cache. DeepSeek also released V4 Flash (284B total/13B active) at $0.14/$0.28 per million tokens, which delivers SWE-Bench performance within 1.6 points of Pro at 12x lower cost.

The model is not without significant caveats. It trails closed frontier models substantially on factual knowledge retrieval (SimpleQA 57.9% vs Gemini's 75.6%) and cross-domain reasoning (Humanity's Last Exam 37.7% vs 44.4%). The 94% hallucination rate means the model nearly always generates a response even when it lacks the underlying knowledge. Political censorship is embedded in the training weights themselves, affecting both hosted and self-hosted deployments by default. For teams building production agents, V4 Pro represents the strongest open-weights option available, but requires careful evaluation against workload-specific requirements and compliance constraints.

분석 생성일: 2026-05-23